Tag: Statistics

The Scottish Exam Results: The real losers are last year’s cohort, and the next!

Now they’ve had a day to do some basic analysis of the Scottish exam results the newspapers have had a chance to put their spin on the story – and the narrative runs something like this:

First narrative – ‘Scottish pupils have had their teacher predicted grades lowered by the qualifications authority’.

Second narrative: – Poor Scottish pupils have had their teacher predicted grades lowered more than rich pupils.

Sources

Links to both the above are at the end of this article

This makes for a great story, but I think they might be misleading. As far as I can see, this year’s National Five Scottish students have done better than they would, on average, had they sat the exams.

If you compare the previous years’ results with the teacher predicted grades you get to see how exaggerated those predictions were…..

A comparison of previous year’s results with teacher predicted grades and the actual downward-adjusted grades

All of the data above is from the articles linked below – NB the blue column for the least and most deprived clusters is only 2019 data, A-C pass rate, and the exam results I’m looking are the National 5s, equivalent to the English GCSE.

What’s really going on?

  1. Teachers in Scotland grossly inflated the predicted grades of their pupils, by 10% compared to previous years on average.
  2. They exaggerated the results of the poorest students more than for rich students (bloody left-wing teachers that is!)
  3. The exam authorities modified the results downards, but the results received are still much better than the previous years, showing an improvement.
  4. The poorest students have improved dramatically.

Analysis

It’s highly unlikely that this bunch of students is hyper-successful compared to previous years, so thus unlikely we would have seen an increase in 10% points in the pass rate.

I think the real thing to keep in mind here is what really goes on in exams – pupils sit them, they are marked, and then stats magic is done on them so we end up with a similar amount of passes and grades distribution to the previous years – so it’s hard-wired into exams that little is going to change year on year.

That’s what we’re seeing here – the exam board adjusting to fit the results in with business as usual, but they’ve had to compromise with those optimistic teachers trying to game the system, and as a result, excuse the pun, this year’s Scottish students have done very well, especiallly the poor.

The students who should be angry are last year’s – they’ve lost out relative to this years, next year’s probably too, and those poor mugs actually had to sit their exams, and didn’t get four months off school!

This probably won’t be the way it’s spun in the media – it’s easy enough to find a few students a parents with individual axes to grind, against the overall trend of the 2020 cohort doing very nicely, thank you teachers!

Sources

The Scottish Sun

BBC News

Social class and educational achievement statistics

The relationship between social class and educational achievement is one of the main topics within the sociology of education at A level.

The problem is, the government does not routinely collect statistics on the relationship between social class and educational achievement!

Instead, we have to reply on statistics which look at the relationship between household income and educational achievement, rather than the relationship between social class and educational achievement.

Household income is related to social class, but income alone does not tell us exactly which social class someone is from. Some parents might work in traditionally ‘working-class’ jobs which could be very well paid, such as the building trades; while other parents might be earning a limited amount of money working part-time in traditionally middle-class jobs – as private music teachers for example.

Also, income does not necessarily tell us about the cultural aspects of class – how well educated parents are or how much social and cultural capital they have, for example.

Thus you must remember that household income indicators are only proxies for social class, they may not show us precisely what a child’s social class background is.

Two sources we might use to to examine the relationship between social class and educational achievement are:

  • Free School Meal (FSM) achievement rates compare to non FSM achievement rates
  • Data on independent school results compared to government schools results.

The Achievement of Pupils Eligible for Free School Meals

Three is a 13.7% achievement gap in the ‘attainment 8’ scores of pupils eligible for Free School Meals compared to non-FSM pupils

In 2019 parents in households with a gross annual income of no more than £16190 were entitled to claim for Free School Meals. (Source).

This means that approximately the poorest 1/6th of households are eligible, so the above statistics are comparing the results of children from the poorest 1/6th of households with the richest 5/6ths all lumped into one.

One limitation with the above statistics is that if you were to stretch this comparison out and compare the poorest 1/6th with the next poorest 1/6th and so on up to the riches 1/6th, you would probably see much starker differences.

Independent School Results Compared to State Schools

If we look at the top 10 independent school results compared to the top 10 state schools, we see quite a difference in results.

In order to be able to pay the fees to get your children into an independent school, you have to be comfortably in the top 10% of households. There are a few scholarships for pupils from poorer households, but not in significant numbers!

Top 10 independent schools

Top 10 state schools

You can see a clear 8-9% difference in achievement in favour of the fee-paying independent schools.

One advantage of the above stats is that it’s much more likely that you’re seeing the solidly upper middle class in these schools, rather than this just being about income.

However, we are only talking about the the top 5-10% of the social class scale, we are not able to make social class comparisons more broadly.

Conclusions

If we use the above data, we can see there is a drastic difference in the achievement rates at the very top and the very bottom of the household income scales.

IF we think household income is a valid indicator of social class, we can also say there are huge social class differences in educational achievement based on the above statistics.

However, we don’t have systematic, annual data on the relationship between the vast majority of middle income households and educational achievement.

Sources

DFE Education Statistics

How does Educational Achievement Vary by Ethnicity?

a look at how GCSE, A-level and degree results vary by ethnic group in England and Wales.

The Department for Education makes it very easy to access statistics on educational achievement. Below I summarise some of the recent trends in educational achievement in England and Wales by ethnicity and offer some commentary on what I think needs explaining, and some thoughts on the limitations of these statistics

Average attainment 8 Score by Ethnic group 2018

Attainment 8 is a way of representing all GCSE results as a single percentage!

The average score for all ethnic groups together was 46.5/90. It’s no surprise to find this is very close to the ‘White British group as White British children still make up the vast majority of school children.

To my mind the headline figures from the above statistics are as follows

  • White, Pakistani and Black African children have results very close to the national average of 46.5, and Bangladeshi children achieved 3% higher. All of these figures are quite close together and so nothing really needs explaining for these broad groups.
  • Chinese children achieve 18% higher than the national average
  • Indian children 10% higher
  • Black Caribbean children underachieve by about 7% points
  • Irish Traveler and Gypsy Roma children have the worst underacheivement levels with 18% and 22% respectively.

So what needs explaining from the above is why Chinese and Indian children do so well, and why Black Carribean children underachieve, and why Irish traveler and Gypsy Roma children do so badly.

In terms of impact of research it’s probably worth focusing on Chinese, Indian and Black Caribbean children because there are many more of these than of the last two ethnic groups.

A final point to note about these statistics is that it doesn’t seem useful to lump together ‘Black’ and Asian’ students because there are SIGNIFICANT differences in the achievement rates within these groups.

Educational Achievement (attainment 8) by Free School Meals and Ethnicity

If we look at GCSE results by free school meal eligibility (roughly the poorest sixth of children) we see that ethnicity still has an independent effect on achievement – the pattern is broadly the same as in the chart above, but with the following two differences:

  • No free school meal children (roughly the wealthiest 5/6ths of children) move closer together slightly.
  • For the FSM groups, white and mixed children are now the lowest achievers, suggesting that poor white and mixed kids do comparatively worse than poor kids from all other ethnic groups.

NB – I think the DFE here is doing a cunning job of disguising the fact that ‘income’ has a larger affect on results than ethnicity – we are seeing here the poorest 1/6th (FSM) compared to the richest 5/6ths (No FSM). If we were to stretch this out and compare just the poorest 1/6th (which we’ve got) to the richest 1/6th my guess would be that you’d find very similar levels of achievement across what would be the upper middle classes for all ethnic groups.

Statistics on Participation in Further Education

This demonstrates a long standing trend – that ethnic minorities (Black and Asian) students are more likely to carry on into further education compared to white students. This should mean that you’ll see a higher proportion of white kids starting work based apprenticeships.

NB – making comparisons to the overall population is a bit misleading as the age profile for ethnic minorities tends to be younger.

Students achieving at least 3 As at A-level

The overall average is 12.9%.

These are quite interesting.

  • Huge ‘over-achievement’ by Chinese kids – 22.5%
  • Indian kids do slightly better than average at 15%
  • Signficant underachievement for Pakistani and Bangladeshi kids – around 7%
  • Terrible underachievement for Black African and Caribbean kids at 5.6% and 3.5% respectively.
  • The source notes that the Irish Traveler population is only 7 people, so one can’t generalize, still, at least it busts a few stereotypes!

These stats show something of an exaggeration of what we saw at GCSE.

I put these stats in the ‘interesting but not that useful’ category – I’d rather see the percentages for high grades or A-C grades to make these a bit more representative.

Degree results by ethnicity

Surprisingly, we see white students gaining significant ground on ethnic minority students with 30.9% of white students gaining a first class degree (*).

Black students in comparison come crashing down to just 14% of first class degrees.

These kind of differences – from similar GCSE results for Black and White students to such different A-level and degree results need further investigation.

(1)The education statistics above form part of the government’s Ethnicity Facts and Figures series, you can check out a wider range of statistical evidence on ethnicity and life chances by clicking here. (As always, remember to be critical of the limitations of these statistics!).

(*) WTF – 30% – sorry kids, but a lot of those first class degrees are probably down to grade inflation, which in turn is probably down to the fact that students are now paying £30 K for yer for their degrees.

Are police officers really 100 times less criminal than the general population?

189 police officers have been convicted across 12 police forces in England in Wales in five years since 2013 , according to a recent FOI request (source: The Telegraph). This equates to just 37.8 police officer convictions each year.

According to Full Fact, there were 126, 300 total police officers in England and Wales in March 2019.

This gives us a police officer conviction rate of 0.03% per year – that is to say that 0.03% of police officers are convicted of a crime each year.

1.38 million people in the general population were prosecuted in the year (CJS Stats, 2018)

A very rough estimate for the number of adults in England and Wales is around 50 million, so this gives us a rough adult conviction rate of 2.76 per year.

This means the Police officer conviction rate is 100 times less than that for the population as a whole.

How accurate are these statistics?

Personally I’m sceptical about the police officer conviction rate.

Despite the fact that the police probably are less likely to commit crime – I mean it kind of goes with the job, not committing crime, and then there’s the embarrassment of getting caught even if you are criminally inclined, which I imagine would be a further deterrent, I still think there’s a lot of criminal police officers whose crimes are just not getting detected.

I imagine you’d be less likely to be suspected of a crime – I mean the police themselves aren’t going to get stopped and searched are they?

Then there’s the fact that prosecutors might be more reluctant to prosecute police because it makes the system look flawed.

Then of course there’s all those things which won’t be defined as criminal because it’s the police doing them in the line of duty – such as speeding and violence, and drug possession come to think of it.

Visualizing the numbers of different types of school in the UK

how useful are different types of data visualization for displaying information about the number of schools in the UK, and how they are changing over time?

I’ve just been experimenting with different ways of visualizing the relative number of different types of school in the UK, and how these school types are changing over time.

The main school types I’ve focused on are:

  • nurseries
  • primary
  • middle (hardly any of these!)
  • secondary
  • non-maintained schools
  • special schools
  • pupil referral units.

You can view all of my visualizations on my Tableau profile. If you click on any of the images below, they will take you to the live vizes on Tableau.

The data comes from DFE’s Education and Training Stats 2018.

Tree Map of School Types in the UK

I think the Tree Map is the best way of providing an overview of the different types of school in the UK. As you can see from the tree map below, there are A LOT more primary schools than secondary schools in the UK, and primary and secondary together (unsurprisingly) account for most of the schools in the UK. In contrast, there are hardly any Pupil Referral Units.

NB – make sure you’ve only got one year box checked, otherwise you get totals of the years checked, which is pointless!

The limitations of the Tree Map Visualization

Although (IMO) this type of viz provides the best ‘frozen in time’ overview, it doesn’t actually allow you to make comparisons across time very easily, because you can only see the distribution for one year at a time.

One could overcome this by having two or more tree maps on display at once, but this still wouldn’t make for easy comparison as the layout of the boxes is likely to change as the data changes. To show changes over time effectively you need different types of visualization:

Changing schools viz 1

This is the most basic type of viz showing changes over time…

line chart showing changes to school types in the UK 2012-2016

Changing schools viz 2

I think viz 2 is much better as the fill gives you a much more immediate impression of how many of each type of school there is.

Changing schools viz 3

Given the relatively small amount of data for this particular viz I think this works quite nicely too, gives you a bit more of a sense of the relative numbers and much better for highlighting smaller numbers…

How useful are these visualizations?

As they stand they only show us a very general level of information, with no granularity. All of these vizes could be much more useful if you could ‘drill down’ into the data to see how the stats vary by England, Wales and Scotland.

Also, these have been designed as only the first stage in a story which also focus in on pupil numbers in different school types (also in relation to pupil types – male/ female etc.), and teacher numbers and teacher-student ratios.

There’s more to come!

Does Prison Work? The Stats suggest not!

What can prison population statistics tell us about Crime Control in the UK?  Is Prison an effective strategy for controlling crime?

These are questions that should be of interest to any student studying the Crime and Deviance option within A-level sociology.

Scotland, England and Wales have high prison populations 

Prison population england.PNG

In England and Wales we lock up 40% more people than in France and almost twice as many people as they do in Germany, which are broadly comparable countries.

Yet there is no link between the prison population and levels of crime 

prison population and crime rate.PNG

  • England and wales have seen a rising prison population and a rising then a rapidly falling crime rate
  • Finland has seen a declining prison population and a rising and then a gradually declining crime rate.
  • Canada has seen a broadly level prison population and yet a relatively stable crime rate.

Most people are serving short sentences for non-violent offences 

what people are sentenced for.PNG

Nearly 70% of the prison population are in for non-violent offences – which means that 30% are in for violent offences. In those prisons where the two populations are mixed, this must be awful for some of those non-violent offenders.

People are getting sentenced for longer 

long sentences for serious offences.PNG

I’m not sure what’s underlying this rise in more serious offences …. the most obvious long-sentence crime of murder has decreased in recent years, so maybe this is for violent gang related and terrorist related crimes which involve in harm rather than death ? Something to research further!

Does Prison work?

In short, if controlling crime is what you hope to achieve, then no it doesn’t because nearly 50% of those sent to prison are recalled within 1 year of being released.

reoffending rates England 2019.PNG

However, there are more reasons why you might want to lock people up other than just rehabilitating them and preventing future offending – there is an argument that they just deserve to be punished whether they reoffend or not.

How do community service orders and suspended sentences compare to prison?

it seems that both of these are more effective at preventing reoffending, but the difference isn’t that great:

  • 63% of people who serve sentences of less than 12 months reoffend compared to
  • 56% of those who receive community orders and compared to
  • 54% of those who receive suspended sentences.

reoffending community service compared prison.PNG

HOWEVER, this may be due to the fact that those avoiding jail have different circumstances and/ or different characters to those who do go to jail – they might just be the kinds of people less likely to reoffend already!

Conclusions 

Overall these prison statistics suggest that while we like to lock people up in England and Wales, there is little evidence that doing so prevents crime.

Maybe we should be looking for cheaper and more effective solutions – such as early intervention (initially expensive but cheaper than several years in and out of jail), or public shaming for example?

Sources 

This post is based on data taken from ‘Prison the facts, Summer 2019‘, published by the Prison Reform Trust.

How many people are in poverty in the UK?

The easy answer is to say around 22% of the population, roughly 14 million people. The long answer starts with the sentence ‘it depends on how you define and measure poverty’, in which case you get various different statistics on the poverty rate.

Statistics on poverty in the UK

According to the Social Metrics Foundation, which seems to be endorsed by the Joseph Rowntree Foundation….

  • 22% of the UK population are in poverty, equivalent to 14.2 million people: 8.4 million working-age adults; 4.5 million children; and 1.4 million pension age adults. Source: The Social Metrics Foundation, 2018.
  • 1% of the total UK population (7. 7 million people) live in persistent poverty. Source: The Social Metrics Foundation, 2018.

This definition of poverty is broader than any previous definition because:

  • It takes account of all material resources not just incomes. For instance, this means including an assessment of the available assets that families have; •
  • It takes into accounts the inescapable costs that some families face, which make them more likely than others to experience poverty, such as the extra costs of disability, and costs of childcare and rental and mortgage costs; •
  • It automatically defines anyone who is ‘sleeping rough’ as being in poverty.

However, it also sets the relative poverty line at 55% of median income rather than 60^ of median income (as the government has done for many years), seemingly because to keep it at 60% while making all of the other changes above would put too many people in poverty?!? See page 63 of the report for more details:

poverty rate UK 2018.png

 

According to the Government’s own data:

  • 16% of UK households were in relative low income households (before housing costs)
  • 22% of UK households were in relative low income households (after housing costs).

Relative low income households have an income of less than 60% of median household income (equivalised), which is equivalent to £296 per week (or approximately £1000 per month). Source: Households Below Average Income, published March 2018.

Households in poverty UK.png

7.3% of the UK population (4.6 million people) are in persistent poverty. This study defines ‘persistent poverty as being in a relative low income household (using the BHAI definition of this) consistently for 3 years. Source: Persistent Poverty in the UK and the EU: 2015.

Which of these is the most valid measurement of poverty?

You’ll notice that there’s some different between these figures, especially between the Social Metric Commissions’ persistent poverty rate and the ONS’ poverty rate – 12% compared to 7%, so it really matters which of these is the most valid!

Given that the Social Metrics Commission’s definition was agreed by a large panel of people, which included government representation, I’m going to say the SMC’s definition/ measurement is the most valid.

Whatever measurement you use, poverty statistics are a terrific example of how statistics are socially constructed.

 

 

The relationship between ethnicity and religion in the UK

According to the 2011 UK census, the religious breakdown of England and Wales was as follows:

  • Christian – 59%
  • No religion – 25%
  • Muslim – 5%
  • Hindu – 1.5%
  • Sikh, Jewish, Buddhist, all <1%

The relationship between ethnicity and religion

  • Christianity is a predominately White religion, especially the Anglican church
  • African forms of Christian spirituality have increased dramatically in the last two decades. Pentecostal Churches are predominately attended by British Africans and African-Caribbeans.
  • Sikhs and Hindus are predominantly of Indian Heritage
  • British Muslims are predominately of Pakistani Heritage, although there is considerable ethnic diversity within British Islam
  • There is some evidence that African-Caribbeans are more likely to be involved in sects such as the Seventh-Day Adventists and the Jehovah’s Witnesses.

Ethnic minorities tend to be more religious than White Britons 

  • Only 32% of adults who reported being Christian said they practiced their religion regularly. This compares to 80% of Muslims and 2/3rds of Hindus, Sikhs and Jews
  • Black Christians are 3 times more likely to attend church than White Christians (English Church Census, 2005)
  • Muslims, Hindus and Black Christians see religion as more central to their identity than White Christians. O’Beirne 2004 found that:
    • Asians, especially Muslims ranked religion and family equally as markers of identity
    • African-Caribbeans and Black-Africans ranked religion as the third most important factor in their lives.
    • White Christians rarely ranked religion as central to their identity.

The relationship between religion and social class

The relationship between social class and religion is not straightforward: the middle classes are, in general, more likely to attend church, but they are also less likely to believe in God and more likely to be atheists and join both world affirming and world rejecting NRMs.

The working classes are less likely to attend church, yet more likely to believe in God than the middle classes. There are also certain denominations and even sects which might appeal specifically to the working classes: such as Methodism, for example.

Church attendance and social class

The ‘middle classes’ have higher rates of church attendance than the ‘working classes’

  • A 2015 YouGov survey of 7000 adults found that 62% of regular church goers were middle class and 38% working class.
  • The same 2015 survey found that twice as many married working class men had never attended church compared to middle class men (17% compared to 9%).
  • Voas and Watt (2014) conducted research on behalf of the Church of England and made three observations not directly about social class, but relevant to it. Firstly, church attendance is higher in rural areas compared to urban areas. Secondly, church attendance is higher in the South of England compared to the North. Thirdly, they noted growth in church attendance in areas which had high performing church primary and secondary schools. All of these indicators suggest higher church attendance in middle class compared to working class areas.
  • Ashworth and Farthing (2007) found that, for both sexes, those in middle class jobs had above average levels of church attendance. Conversely, those in skilled, semi-skilled and unskilled working class jobs had below average church attendance. Welfare recipients had the lowest levels of church attendance.

Religious belief and social class

  • A 2016 YouGov Survey revealed that 48% of those in social grades ABC1 described themselves as ‘Atheist’ compared to 42% of those in social grades C2ED.

  • A 2013 review of >60 research studies on the relationship between IQ and religiosity found that people with higher IQs are more likely to be atheists. (High IQs are correlated with higher levels of education and higher social class).
  • Lawes (2009) found that ‘lifelong theists’ disproportionately come from unskilled and semi-skilled manual backgrounds, and were less likely to have academic qualifications. Conversely, lifelong atheists disproportionately come from higher professional and managerial backgrounds, and are more likely to have experienced higher education.

NB – It’s worth noting how this contradicts what’s above in terms of church attendance

Social class, religion and deprivation 

There is some evidence that those suffering deprivation (the lower social classes) are more likely to turn to religion…..

  • Churches in deprived inner city areas tend to have higher rates of attendance.
  • Methodist, Pentacostal and Baptist denominations  tend to be more working class.
  • Catholic Churches are more likely to attract Irish, Polish and African immigrants who have typically experienced higher levels of deprivation.

New Religious Movements and social class

As a general rule, the middle classes are more attracted to both World Affirming NRMs (and the New Age Movement), and World Rejecting NRMs, at least according to Eileen Barker’s classic study of ‘The Moonies’.

Problems with identifying the relationship between religion and social class

  1. Andrew Mckinnon notes that there has been a ‘dearth’ of research on the relationship between religion and social class, meaning there is something of a data gap.
  2. Because of the above, we are often stuck with relying on indicators which might not actually measure social class.
  3. Even if the data suggests that church attendance and belief are higher among the middle classes, this doesn’t necessarily mean the middle classes are actually more religious. They may just be attending church to keep up appearances or to get their children into the local church school (which tend to have high academic performance); or they  may feel under more social pressure to state they are religious than the working classes

Sources: 

Chapman et al, as well as the good ole’ t’internet.

Is alcohol really that bad for your health?

The new ‘safe’ level of alcohol consumption should be none, at least according to a recent study into the health risks of alcohol published by the The Lancet.

This contradicts the current official government guidelines on the ‘safe’ level of drinking: currently around 14 units per week for women, and 21 for men.

The findings of this research study were widely reported in the mainstream media:

  • The Daily Mail reported that ‘just one glass of wine a day increases your risk of various cancers’.
  • Even The Independent reported that ‘the idea that one or two drinks a day is good for you is a myth’.

alcohol health statistics.png

But what are the actual statistical risks of different levels of alcohol consumption?

The actual risk of developing a drink related alcohol problem for different levels of drinking are as follows:

  • No drinks a day = 914/ 100 000 people
  • One drink a day = 918/ 100 000 people
  • Two drinks a day = 977/ 100 000 people

I took the liberty of putting this into graph form to illustrate the relative risks: blue shows the proportion of people who will develop alcohol related problems!

alcohol health risks

This means that statistically, there is only a 0.5 % greater risk of developing an alcohol related illness if you have one drink a day compared to no drinks, which hardly sounds significant!

Meanwhile, there is a greater increase in risk if you have two compared to 1 drink a day, which suggests the government guidelines have got this about right!

(NB, despite the headlines, The BBC and Sky did a reasonable job of reporting the actual stats!)

So why did some news papers report these findings in a limited way?

This could be a classic example of News Values determining how an event gets reported: it’s much more shocking to report that the government has got its advice wrong and that really there is no safe level of drinking!

Or it could be that these newspapers feel as though they’ve got a social policy duty to the general public… even if there is only a slight increased risk from alcohol consumption, maybe they feel duty bound to report it in such a way to nudge behaviour in a more healthy direction.

In terms of why some newspapers did a better job of reporting the actual findings: it could be that these are the papers who rely on advertising revenue from drinks companies? Maybe the Mail and the Independent don’t get paid by drinks companies, whereas Sky does>?

This post will also be published to the steem blockchain: where you can get rewarded in crypto currency for posting, liking, commenting and so on!